What if the NBA and WNBA used the Elam Ending?
The Elam Ending turns the game clock off near the end and plays to a target score instead. The Basketball Tournament (TBT) version used here switches on at the first dead ball with under four minutes left in the fourth quarter, and the first team to reach the leading score plus 8 wins. This study asks what that rule would have done to the last three seasons of both leagues: 4,866 games, every regular-season and playoff game of NBA 2023-24 to 2025-26 and WNBA 2024 to 2026. The short answer: it would change the winner of a few close games, barely move the odds of the team ahead, and cut the slowest minutes in basketball.
Three ways to answer something unobservable
Nobody knows exactly how these games would have gone without a clock, because teams play the end of a game to the clock: they foul to stop it, hold for the last shot, and call timeouts to set up a play. So the question gets three answers, each with an honest claim attached.
The first is a literal replay. Each game's real play-by-play is read from the trigger on, and the first team whose real score reaches the target wins. That keeps every clock-driven decision the teams actually made, so it is a bound, not a forecast.
The second is a model of the race to the target. From the score and possession at the trigger, it computes the exact chance that each team gets there first, using how often possessions score 0, 1, 2, 3, or 4 points when nobody is playing to the clock. Before it is used on endings, it is checked against real races to "leader + 8" that happen inside games, in the second and third quarters.
The third is an audit of the endings the rule would remove: overtime, fouls to stop the clock, the free-throw share of late points, and how long the last two minutes really take.
Every result was computed twice, in Python and in R, and the two agree on every cell. Final scores were checked against the league's game logs (all 4,866 match) and late-game scores against a second play-by-play feed.
A few close games change hands
The replay stops each game the moment a team reaches the target, so a real comeback after that moment no longer counts. The game with the latest such finish shows how that changes a winner: the Lakers reached the target first, and the Bulls won the real game at the buzzer.
Under a literal replay, 3.0% of NBA regular-season games and 3.0% of WNBA regular-season games crown the other team. Most of them were close at the trigger: 16% of NBA games tied at the trigger change winner, and no game with a lead of more than 15 does.
The replay cannot decide every game. In 38% of NBA regular-season games neither team reaches the target before the real final buzzer. Those games are mostly lopsided: 57% of NBA games led by 16 or more at the trigger end that way, against 13% of one-point games. The race model below covers them.
The team ahead keeps almost exactly its odds
The race model reads the same game from its trigger state instead of from what happened next. It made the Lakers clear favorites to reach the target first, and in the replay they did.
On the 9,092 real mid-game races, the model said the favorite would get there first 86.8% of the time; it did 86.2% of the time. No range of predictions was off by more than 1.3 points, and its Brier score was 0.098, against 0.119 for always guessing the overall rate.
Applied to the endings, it says the rule would barely change how often the team ahead at the trigger wins: 90.7% of NBA regular-season games under the clock, 90.6% under Elam, and 90.0% against 89.3% in the WNBA. The clearest shift is at big leads. NBA teams up 11 to 15 won 99.7% of real games, with a 95% interval of 98.9% to 99.9%, and the model gives them 98.4%: with no clock to run out, a trailing team always gets one more chance.
The playoffs show bigger gaps, but they rest on 313 games with a leader at the trigger, too few to read much into.
The slowest minutes in basketball
The clock shapes the end of a real game in ways the rule would remove.
| Overtime | Free-throw share of points, last 4:00 and overtime (Q1-Q3) | Trailing team commits a take foul, last 2:00 | Real minutes for the last 2:00, close games (others) | |
|---|---|---|---|---|
| NBA regular season | 4.7% | 23% (14%) | 26% of games | 10.2 (4.6) |
| WNBA regular season | 2.9% | 28% (17%) | 25% of games | 11.3 (5.0) |
Free throws make up 23% of NBA points in the last four minutes and overtime, against 14% in the first three quarters. A trailing team commits a take foul, a deliberate foul to stop the clock, in the last two minutes of 26% of NBA games. And the last two minutes of a close NBA game take a median of 10.2 real minutes, more than twice the 4.6 of other games; NBA games separated by more than 20 at 2:00 take 3.5.
What this does not show
The literal replay keeps the clock strategy the real teams used, and the race model assumes possessions score the way they do when nobody watches the clock. Real Elam endings would sit somewhere else: teams would stop fouling and holding, and might shoot differently with a target in sight. The model also treats each possession as independent and gives each team its league's or its own season's scoring rates, with no adjustment for who is on the floor. The NBA's real target-score games are All-Star exhibitions, too few and too unlike competitive play to test it on. The 2026 WNBA Finals were still being played when the data was first pulled; they were added once the series ended.
Method, code, and verification
The full technical write-up (data provenance, model specification, validation gates, and stated limitations) is the README. The analysis is implemented independently in R and Python and reconciled to numeric tolerance; run_checks.sh replays every gate.